Autonomous Navigation of MAVs in Unknown Cluttered Environments (1906.08839v3)

Abstract: This paper presents an autonomous navigation framework for reaching a goal in unknown 3D cluttered environments. The framework consists of three main components. First, a computationally efficient method for mapping the environment from the disparity measurements obtained from a depth sensor. Second, a stochastic method to generate a path to a given goal, taking into account field of view constraints on the space that is assumed to be safe for navigation. Third, a fast method for the online generation of motion plans, taking into account the robot's dynamic constraints, model, and environmental uncertainty and disturbances. To highlight the contribution with respect to the available literature, we provide a qualitative and quantitative comparison with the state of the art methods for reaching a goal and for exploration in unknown environments, showing the superior performance of our approach. To illustrate the effectiveness of the proposed framework, we present experiments in multiple indoors and outdoors environments running the algorithm fully on board and in real-time, using a robotic platform based on the Intel Ready to Fly drone kit, which represents the implementation in the most frugal platform for navigation in unknown cluttered environments demonstrated to date. Open source code is available at:~\url{https://github.com/IntelLabs/autonomousmavs}. The video of the experimental results can be found at~\url{https://youtu.be/Wq0e7vF6nZM}.

• The paper introduces a comprehensive navigation framework for MAVs that integrates efficient RGB-D mapping with stochastic path planning and rapid online motion planning.
• It employs a linear octree structure with a tailored stochastic method to achieve real-time performance at 33Hz using onboard processing.
• The framework outperforms traditional optimization-based methods, offering practical advantages for applications in search and rescue, industrial inspection, and environmental monitoring.

An Autonomous Navigation Framework for MAVs in 3D Cluttered Environments

This paper introduces a comprehensive framework for the autonomous navigation of Multirotor Micro Aerial Vehicles (MAVs) in unknown and cluttered three-dimensional environments. The proposed framework is built upon three primary components: efficient environment mapping, stochastic path planning, and rapid online motion planning, each designed to enhance the capability of MAVs to navigate and avoid collisions in complex environments.

Framework Components and Methodology

1. Environment Mapping: The framework employs a computationally efficient method for environment mapping using disparity measurements from RGB-D cameras. This technique facilitates the creation of a sparse yet effective representation of the environment by leveraging a linear octree data structure. The octree structure aids in efficient collision detection, which is crucial for real-time navigation.
2. Stochastic Path Planning: Path planning is achieved through a stochastic method taking into account the MAV's field of view and navigable space. The method is tailored to work within the constraints of a three-dimensional space without necessitating high-order dynamic calculations, thereby optimizing computational efficiency.
3. Online Motion Planning: The online motion planning component addresses the dynamic constraints and uncertainties inherent in the environment, allowing for rapid adjustment to the planned paths. The framework eschews the need for optimization solvers, thereby making it suitable for operations on platforms with limited computational power.

Experimental Evaluation

The efficacy of the proposed framework is demonstrated through extensive experiments conducted in various indoor and outdoor environments. Utilizing a robotic platform based on the Intel Ready to Fly drone kit, the experiments showcased the system's capability to achieve real-time autonomous navigation using only onboard processing resources. Notably, the experimental results highlight the system's ability to operate at the camera's frame rate of 33Hz, surpassing other state-of-the-art methods in terms of computational simplicity and onboard resource efficiency.

Qualitative and Quantitative Comparisons

The framework is systematically compared with existing approaches concerning several crucial aspects of autonomous navigation, including navigation in the field of view, dynamic and positional constraints, and the ability to escape local minima or "pockets." While many existing methods rely heavily on computationally intensive optimization processes, this framework offers a more efficient solution by integrating a complete navigation method that leverages a lightweight motion planning strategy coupled with a robust mapping technique.

Implications and Future Directions

Practically, the framework offers significant advancements in MAV navigation, particularly in scenarios where onboard computational power and energy consumption are constrained. Theoretically, it contributes to ongoing research in autonomous robotics by providing a new approach that integrates mapping, planning, and control components required for comprehensive autonomous navigation in unknown environments.

Future enhancements could include the integration of trajectory tracking capabilities and the prediction of dynamic obstacles, enabling more sophisticated navigation strategies in ever-changing environments. These developments will further solidify the framework's position as a viable solution for autonomous exploration and task execution in demanding and unstructured environments.

This research propels the capabilities of MAVs in complex navigational tasks, offering a foundation for future studies and practical applications in fields such as search and rescue, industrial inspection, and environmental monitoring.